Test hypothesis 2)

facebook · Mar 10, 2025 · 7202178 · 7202178
1 parent 9807d06
commit 7202178
Show file tree

Hide file tree

Showing 3 changed files with 30 additions and 19 deletions.
diff --git a/db/wal_manager.cc b/db/wal_manager.cc
@@ -12,6 +12,7 @@
 #include <algorithm>
 #include <cinttypes>
 #include <memory>
+#include <thread>
 #include <vector>
 
 #include "db/log_reader.h"
@@ -172,8 +173,9 @@ void WalManager::PurgeObsoleteWALFiles() {
   s = env_->GetChildren(archival_dir, &files);
   if (!s.ok()) {
     IGNORE_STATUS_IF_ERROR(s);
-    ROCKS_LOG_ERROR(db_options_.info_log, "Can't get archive files: %s",
-                    s.ToString().c_str());
+    ROCKS_LOG_ERROR(
+        db_options_.info_log, "Can't get archive files: %s, Thread id: %lu",
+        s.ToString().c_str(), (unsigned long)pthread_self());
     return;
   }
 
@@ -190,17 +192,20 @@ void WalManager::PurgeObsoleteWALFiles() {
         if (!s.ok()) {
           IGNORE_STATUS_IF_ERROR(s);
           ROCKS_LOG_WARN(db_options_.info_log,
-                         "Can't get file mod time: %s: %s", file_path.c_str(),
-                         s.ToString().c_str());
+                         "Can't get file mod time: %s: %s, Thread id: %lu",
+                         file_path.c_str(), s.ToString().c_str(),
+                         (unsigned long)pthread_self());
           continue;
         }
         if (now_seconds - file_m_time > db_options_.WAL_ttl_seconds) {
           s = DeleteDBFile(&db_options_, file_path, archival_dir, false,
                            /*force_fg=*/!wal_in_db_path_);
           if (!s.ok()) {
             IGNORE_STATUS_IF_ERROR(s);
-            ROCKS_LOG_WARN(db_options_.info_log, "Can't delete file: %s: %s",
-                           file_path.c_str(), s.ToString().c_str());
+            ROCKS_LOG_WARN(db_options_.info_log,
+                           "Can't delete file: %s: %s, Thread id: %lu",
+                           file_path.c_str(), s.ToString().c_str(),
+                           (unsigned long)pthread_self());
             continue;
           } else {
             MutexLock l(&read_first_record_cache_mutex_);
@@ -216,8 +221,9 @@ void WalManager::PurgeObsoleteWALFiles() {
         if (!s.ok()) {
           IGNORE_STATUS_IF_ERROR(s);
           ROCKS_LOG_ERROR(db_options_.info_log,
-                          "Unable to get file size: %s: %s", file_path.c_str(),
-                          s.ToString().c_str());
+                          "Unable to get file size: %s: %s, , Thread id: %lu",
+                          file_path.c_str(), s.ToString().c_str(),
+                          (unsigned long)pthread_self());
           return;
         } else {
           if (file_size > 0) {
@@ -229,8 +235,9 @@ void WalManager::PurgeObsoleteWALFiles() {
             if (!s.ok()) {
               IGNORE_STATUS_IF_ERROR(s);
               ROCKS_LOG_WARN(db_options_.info_log,
-                             "Unable to delete file: %s: %s", file_path.c_str(),
-                             s.ToString().c_str());
+                             "Unable to delete file: %s: %s, Thread id: %lu",
+                             file_path.c_str(), s.ToString().c_str(),
+                             (unsigned long)pthread_self());
               continue;
             } else {
               MutexLock l(&read_first_record_cache_mutex_);
@@ -259,8 +266,9 @@ void WalManager::PurgeObsoleteWALFiles() {
   if (!s.ok()) {
     IGNORE_STATUS_IF_ERROR(s);
     ROCKS_LOG_WARN(db_options_.info_log,
-                   "Unable to get archived WALs from: %s: %s",
-                   archival_dir.c_str(), s.ToString().c_str());
+                   "Unable to get archived WALs from: %s: %s, Thread id: %lu",
+                   archival_dir.c_str(), s.ToString().c_str(),
+                   (unsigned long)pthread_self());
     files_del_num = 0;
   } else if (files_del_num > archived_logs.size()) {
     ROCKS_LOG_WARN(db_options_.info_log,
@@ -275,8 +283,10 @@ void WalManager::PurgeObsoleteWALFiles() {
                      /*force_fg=*/!wal_in_db_path_);
     if (!s.ok()) {
       IGNORE_STATUS_IF_ERROR(s);
-      ROCKS_LOG_WARN(db_options_.info_log, "Unable to delete file: %s: %s",
-                     file_path.c_str(), s.ToString().c_str());
+      ROCKS_LOG_WARN(db_options_.info_log,
+                     "Unable to delete file: %s: %s, Thread id: %lu",
+                     file_path.c_str(), s.ToString().c_str(),
+                     (unsigned long)pthread_self());
       continue;
     } else {
       MutexLock l(&read_first_record_cache_mutex_);

diff --git a/db_stress_tool/db_stress_shared_state.h b/db_stress_tool/db_stress_shared_state.h
@@ -378,7 +378,10 @@ class SharedState {
   }
 
  private:
-  static void IgnoreReadErrorCallback(void*) { ignore_read_error = true; }
+  static void IgnoreReadErrorCallback(void*) {
+    fprintf(stdout, "DBStress thread id: %lu\n", (unsigned long)pthread_self());
+    ignore_read_error = true;
+  }
 
   // Pick random keys in each column family that will not experience overwrite.
   std::unordered_set<int64_t> GenerateNoOverwriteIds() const {

diff --git a/db_stress_tool/no_batched_ops_stress.cc b/db_stress_tool/no_batched_ops_stress.cc
@@ -1685,10 +1685,6 @@ class NonBatchedOpsStressTest : public StressTest {
       s = iter->status();
     }
 
-    if (ro_copy.auto_refresh_iterator_with_snapshot) {
-      std::this_thread::sleep_for(std::chrono::microseconds(3 * 1000 * 1000));
-    }
-
     int injected_error_count = 0;
     if (fault_fs_guard) {
       injected_error_count = GetMinInjectedErrorCount(
@@ -1701,6 +1697,8 @@ class NonBatchedOpsStressTest : public StressTest {
         // Grab mutex so multiple thread don't try to print the
         // stack trace at the same time
         MutexLock l(thread->shared->GetMutex());
+        fprintf(stderr, "DBStress thread id: %lu\n",
+                (unsigned long)pthread_self());
         fprintf(stderr, "Didn't get expected error from PrefixScan\n");
         fprintf(stderr, "Callstack that injected the fault\n");
         fault_fs_guard->PrintInjectedThreadLocalErrorBacktrace(